If you’ve ever wished ChatGPT had taken off sooner, it looks like OpenAI has answered your prayers.
AI Labs has rolled out a new mode called Ultrafast. It is said to be designed to significantly accelerate the work speed of the latest and most powerful model, GPT-5.6 Sol.
The company says Ultrafast runs 14 times faster than standard processing and can output up to 750 output tokens (tokens representing individual pieces of text that the LLM generates when interacting with a human) per second.
“Until now, real-time speeds have typically been achieved by choosing smaller or more specialized models,” the company said in a blog post Thursday. “Ultrahigh speeds represent progress in a new direction: increasing the amount of useful work per second.”
OpenAI competitors, such as Anthropic, have similarly released accelerated versions of their models. Claude has a fast mode, but it doesn’t offer the speed that OpenAI offers here.
OpenAI suggests that this highly functional version of GPT-5.6 Sol can be deployed across a variety of enterprise workflows, particularly incident response, customer service and support, financial market analysis, e-commerce, and other related areas.
Currently released in preview, Ultrafast is made possible by a partnership between OpenAI and chipmaker Cerebras. Currently, that preview is only available to some customers, but OpenAI says it will expand access to this feature “as capacity increases.”
